Sharpening and Resize Observations

Andy_L wrote on 7/14/2010, 5:43 PM
I've been running some tests applying the same Unsharp Mask Fx at various points in the signal to a clip that will be resized on render. The typical scenario would be HD to SD, but this would also be a common conversion for web video as well.

The test is simple: I export stills from the end result and look at them in Photoshop to see if I can detect any differences.

The results have been somewhat surprising.

As far as I can tell, it makes no difference in the signal path where you sharpen--the end result is always the same.

That is, sharpening a clip (or even media in the project media box!) gives you the same result as sharpening at the output bus.

In other words, Vegas always sharpens after resizing the video.

How about if you sharpen a clip, set project properties to the target resolution, and then nest that project in a new project?

Same result: vegas still applies sharpening after the downresolution, rather than before it.

The good news on this is that in nearly all situations, you do want to sharpen after resizing. Bad news is that Vegas seems to be overriding its own signal path in some of these tests, so if you do want to force a sharpen before the resolution is changed (ie, to do a capture sharpen), your only choice as far as I can tell is to physically render out to an intermediate.

One last note: for the SD crowd, sharpening may be a bad idea to begin with, as it boosts high-frequency detail which will then compete with your subject matter and motion for the limited bitrate budget of your MPEG-2 compressor.


VP 9.0e win 7/64

Comments

farss wrote on 7/14/2010, 6:08 PM
Your observations are in line with mine however you may have fallen into a couple of traps:

1) We are dealing with moving images. Moving images have an order of magnitude more complexity than still images. I could probably dig up a white paper from the BBC that explains how too much sharpening can cause the appearance of negative motion.

2) The Nyquist problem. This get way more messy with video and interlaced video in particular where the vertical and horizontal sampling frequncies are not the same. The viewable results include aliasing and line twitter.

I should say upfront that I only understand any of this well enough to know why something is getting messed up and how to avoid it. The underlying theory and maths is way over my head.

My practical approach these days is to edit at native res within a project that matches the media. I then Nest that within a project set to the delivery format and make adjustments to suit. If I have any doubts about what I'm doing I monitor on a CRT based monitor.

I do see some pretty bad aliasing and line twitter problems in OTA broadcasting. I think I can safely say this is not a trivial problem and with the tools that Vegas has we can do a pretty decent job if you invest the time to understand the problems and how to use the tools we have to solve them. Others with a more creative bent than me may understandably think otherwise.

Bob.
musicvid10 wrote on 7/14/2010, 6:15 PM
Andy_L,

Actually, Bob is being a bit modest here. Although he may not understand the math, his past observations about where in the chain to apply the sharpening and enhancements changed my level of attention, if not my approach to everything from resizing to transcoding.

I knew a lot about this (I thought) from my analog editing days, but a fresh look never hurt anyone. Search his old posts and you will find your excellent observations will be both supported and challenged.
Andy_L wrote on 7/15/2010, 7:55 AM
I should say for my tests I was exclusively looking at progressive 720p footage.

The introduction of interlaced footage creates more than a few twists where resizing and sharpening are concerned in Vegas.